EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: dcarr on June 13, 2015, 04:35:45 am
-
If you're using the signal generators in the CMU200 (and probably CRTU) there's a gotcha you should probably know about. Check out the waveform you get when you enable the generator output:
(http://i1277.photobucket.com/albums/y486/digidocs/func_on_zps7yey3218.png)
That's a 3V+ spike it delivers a few ms before activating the generator. This cost me an entire day of debugging on a new radio design--the spike caused the internal registers of the chip to be corrupted. Fun!
David
-
If you're using the signal generators in the CMU200 (and probably CRTU) there's a gotcha you should probably know about. Check out the waveform you get when you enable the generator output:
(http://i1277.photobucket.com/albums/y486/digidocs/func_on_zps7yey3218.png)
That's a 3V+ spike it delivers a few ms before activating the generator.
What exactly are the test parameters (i.e. input frequency, input level, cabling, any devices in the line other than cables, termination on the scope end)?
I did a quick test on my CRTU but I could only get a similar spike (although of much lower voltage) when my scope is in 1Mohms mode (i.e. improperly terminated transmission line), and even then it was pretty rare occurrence. I couldn't see any spikes when in 50ohms mode.
Also note that this is pretty standard behavior of many RF generators. That's why it's standard practice to set the output level to something very low (i.e. -110dBm) before enabling RF, and only after switching RF on changing the output to the desired level.
-
I tested this again to make sure that I hadn't accidentally left the line unterminated.
My setup is RF3 OUT -> 3' BNC cable -> T-adapter on scope with terminator on one end of T fitting.
I tried with levels from +13 to -90dBm and got the same glitch. Maybe your hardware is different than mine?
David
-
I tested this again to make sure that I hadn't accidentally left the line unterminated.
My setup is RF3 OUT -> 3' BNC cable -> T-adapter on scope with terminator on one end of T fitting.
I tried with levels from +13 to -90dBm and got the same glitch.
What frequency? Did you try other frequencies? Did the spikes show up every time, or only after a certain number of RF on/off cycles? Does it show when immediately switching RF back on after switching off, or only when there's some time between off and on?
What's the bandwidth of the cable and adapters?
Maybe your hardware is different than mine?
I only have a CRTU (sold my CMU200) so yes, the hardware is different, but I haven't really seen such behavior on my old CMU200 as well.
But as I said, it's pretty common for many RF generators (even standalone ones) to at least occasionally spike when switching RF on.
It's also possible that something in your test setup causes that spike. Do you have a chance to use an inline terminator on the scope side?
-
If you want to enable a signal with minimal transient issues the normal procedure with a decent (modernish) standalone lab sig gen is to start with the generator set to RF CARRIER OFF and then set the frequency and then set the level. Then set it to CARRIER ON.
This will provide the least risk of transients.
I've not used a CMU200 test set but it is going to be compromised in various ways that will mean the spurious, noise (and transient?) performance won't be as good as a conventional sig gen. Maybe you are expecting too much from a test set although that 3V spike looks bad to me. Could your unit be faulty?
-
If you want to enable a signal with minimal transient issues the normal procedure with a decent (modernish) standalone lab sig gen is to start with the generator set to RF CARRIER OFF and then set the frequency and then set the level. Then set it to CARRIER ON.
Yes, with modern kit, which usually has some kind of transient suppression. However, in general thest practice is RF off, select frequency, set output power to minimum, enable RF and then set the desired output level. Especially on older kit (i.e. HP 8672A) it's essentially the only way to avoid transients.
I've not used a CMU200 test set but it is going to be compromised in various ways that will mean the spurious, noise (and transient?) performance won't be as good as a conventional sig gen. Maybe you are expecting too much from a test set although that 3V spike looks bad to me. Could your unit be faulty?
I haven't seen any such transients on any CMU200 (or CRTU, which shares some hardware with the CMU200) on a properly terminated transmission line. The only time I saw any spikes (although not at that amplitude) when enabling my CRTU's RF output was when it was connected to an 1M scope input, and even then the rate of occurrence was very low. And that was with a cheap-ass BNC lead which today turned out to be defective anyways (BNC connector came off) and probably already was when I did the quick test.
It could very well be that something in his test setup causes this transient.
-
Yes, with modern kit, which usually has some kind of transient suppression. However, in general thest practice is RF off, select frequency, set output power to minimum, enable RF and then set the desired output level. Especially on older kit (i.e. HP 8672A) it's essentially the only way to avoid transients.
The HP8672A is a microwave generator that starts at 2GHz and covers up to 18GHz using multipliers. I'm assuming we are comparing like with like here. Therefore, we should be considering the behaviour of mainstream sig gens that cover LF through to maybe a few GHz.
Yes, with modern kit, which usually has some kind of transient suppression.
As far as I know modern(ish) sig gens that cover LF to maybe a few GHz don't need transient suppression. By modernish I'm going back maybe 35 years and looking at medium to high grade sig gens. Eg Marconi or HP/Agilent or maybe R&S that can be purchased for a few hundred pounds today.
I would hope/expect them all to use a similar tactic where (for CARRIER OFF) the generator keeps the signal running deep inside the generator but isolates the signal from the final stages of the sig gen and selects maximum attenuation in the output attenuators. That would equate to carrier OFF.
Then to turn the carrier ON I assume the sequence would be to turn on the drive to the final stages and allow the ALC to settle to its target level and then release the attenuators to the required attenuation for the required power level. The ALC should settle very quickly as it needs to be very fast to support fast AM modulation modes.
The sig gen knows the ALC target in advance and it knows the attenuator setting it needs after this. So there should be no need to suppress any transients. The whole sequence should take a tiny fraction of a second.
If you do it the way you suggest, you run the slight risk of ALC transients caused by sudden step attenuator changes. When the step attenuator changes a significant step, the ALC needs to set a new target and hit it without glitching/overshoot. This is a known weakness of modern sig gens and some generators offer a (step) 'attenuator hold' feature to allow the user to vary the level over a certain limited range whilst the carrier is already ON if they are worried about transients from the step attenuator change and the associated sudden large change in ALC target.
-
I tested this again to make sure that I hadn't accidentally left the line unterminated.
My setup is RF3 OUT -> 3' BNC cable -> T-adapter on scope with terminator on one end of T fitting.
I tried with levels from +13 to -90dBm and got the same glitch. Maybe your hardware is different than mine?
David
I've never used a CMU200 but looking at the block diagram below, the bit in green is the equivalent of a basic RF sig gen with a programmable attenuator at the output (arrowed in green).
But the signal then has to go through a complex matrix of attenuators, splitters and switches and these have to be rated at a power level much higher than the sig gen can deliver. So maybe these 'extra' switches circled in red are producing switching glitches after the normal attenuator circled in green.
A conventional sig gen would not have the extra (and very complex and very lossy) matrix containing the switches circled in red. I'm guessing that the Pi attenuators circled in blue have to be rated at >50W and will be fan cooled. The switches next to this (in red) probably still have to have a fairly high RF power rating and a side effect may be that you get some switching glitches leaked from the switch control voltages/currents.
But that's just a guess. Maybe you could email/ask a technical rep at R&S?
-
I've seen that problem on some 20 year old R&S sig gens but I can't remember which model it was.